Deep-Sea Treasure, Image version

The Deep-Sea Treasure, Image version is an image state variant of a multi-objective reinforcement learning toy environment, first introduced by P. Vamplew et al. in "Empirical Evaluation Methods for Multi-Objective Reinforcement Learning Algorithms." This task simulates the process of treasure hunting in the deep sea through image inputs, aiming to evaluate and optimize the performance and decision-making capabilities of multi-objective reinforcement learning algorithms in complex visual environments, which holds significant research and application value.